[napetrov]

Summary

Test-only. Two leftovers from #94 (after #252).

tests/
├── test_matrix_rollup.py        assert args.func is cmd_arms_matrix_run      (was args.arms_cmd)
├── test_novelty_cli.py          inspect.unwrap(args.func) is the handler      (was args.novelty_cmd)
└── test_printed_hints_parse.py  + test_no_string_in_the_package_names_the_source_shim

• Hints (item 1): already fixed on intel/main by fix(cli): next-step hints name the dokimi command; drop dead cmd_report #252. No string literal in dokimi/ names
  python cli.py any more. This PR adds the guard: an AST scan of every string literal in the
  package (docstrings excluded) that fails on python cli.py, so a new hint that copies the source
  shim cannot land unnoticed.
• Subparser dest (item 2): the two tests read the argparse dest names (arms_cmd,
  novelty_cmd). They now assert the command that runs. A later move of a subcommand group keeps
  them green.

Production code is unchanged. No changelog fragment: no user-visible behaviour changes. Refs #94.
The numbers do not change.

Evidence

• Before (guard red): the hint reverted to python cli.py in commands/personas.py:73:

  E  AssertionError: hints name the source shim, not `dokimi`:
  E    commands/personas.py:73: 3. Approve: python cli.py personas approve --file
  1 failed
  

  After: on current main, tests/test_printed_hints_parse.py: 6 passed.

• Before (dest rename): dest="arms_cmd" and dest="novelty_cmd" renamed in the source. The old
  asserts fail:

  E  AttributeError: 'Namespace' object has no attribute 'arms_cmd'
  E  AttributeError: 'Namespace' object has no attribute 'novelty_cmd'
  2 failed
  

  After: the same rename, with the new asserts: 2 passed.

• Full suite (pytest -q -n 8): this branch 9716 passed, 6 skipped. intel/main 114cb6c
  9715 passed, 6 skipped. The extra pass is the guard. Lint, mypy, deptry, changed-line coverage and
  the test-quality audit pass under make ci-local.

Merge Danger

Door: two-way. Test-only changes; revert is trivial.

Open PRs: none of the three files is in any open PR's file list (checked on 2026-10-29). PRs #254
(CLI flag registry and snapshot, #94 F1a) and #238 touch CLI code, so rebase them if they land first.

Blast Radius: test suite only. A failure here means a test is wrong, never shipped behaviour.

🤖 Generated with Claude Code

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 93.02%. Comparing base (114cb6c) to head (26e9f43).
❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #283   +/-   ##
=======================================
  Coverage   93.02%   93.02%           
=======================================
  Files         265      265           
  Lines       33464    33464           
  Branches     5610     5610           
=======================================
  Hits        31130    31130           
  Misses       1540     1540           
  Partials      794      794           

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[vlad-perevezentsev]

LGTM

I checked 26e9f43. The parser tests now check which handler runs, and the new guard catches a python cli.py hint anywhere in the package.

What I ran, at 26e9f43:

• make ci-local --offline: 0 failed, 17 passed, 13 skipped (network-only jobs).
• Mutation: an f-string f"run: python cli.py arms run {1}" added to _version.py makes test_no_string_in_the_package_names_the_source_shim fail and name the line. The base has no such string, so the guard is a forward check by design.

Checked, no issues

• Handler assertions: inspect.unwrap(args.func) is handler works because _stops_at_cap uses functools.wraps; matrix-run binds cmd_arms_matrix_run directly.
• Docstring exclusion: only the first statement of a module, class or function is skipped, so strings in code are still scanned, f-string parts included.

[antonwolfy]

LGTM. The handler assertions check the command that runs, and the new guard catches a python cli.py string anywhere in the package.

What I ran, on a Linux review host at 26e9f43:

• make ci-local: 27 passed, 1 failed, 7 skipped. 9718 tests passed, 4 skipped. The one failure is pip-audit, which reports advisories against filelock 3.32.7 in the lock files. This PR is test-only and does not touch them.
• Rename check: renaming dest="arms_cmd" and dest="novelty_cmd" fails the old asserts at the base (2 failed) and keeps the new ones green (91 passed).
• Mutation: wiring novelty probe to cmd_novelty_atoms fails test_every_subcommand_is_registered. Wiring matrix-run to cmd_arms_run fails the matrix parser test.
• Guard mutation: plain strings, f-strings and implicitly concatenated literals containing python cli.py all fail the guard.

Checked, no issues

• Unwrap: _stops_at_cap uses functools.wraps, so inspect.unwrap reaches the handler for analyze, atoms, probe and shorten. view is registered bare.
• Scan root: Path(dokimi.__file__).parent resolves to the package under test, and rglob covers its subdirectories.
• Lint: ruff check passes on the changed tests, and both new imports are used.
• Commits: both are signed off.